Cache the current fault info in the vm struct. This can be queried
by userspace later to help debug UMDs.
Cc: samuel.pitoiset@gmail.com
Reviewed-by: Christian König <christian.koenig@amd.com>
Acked-by: Guchun Chen <guchun.chen@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
When we get a GPU page fault, cache the fault for later
analysis.
Cc: samuel.pitoiset@gmail.com
Reviewed-by: Christian König <christian.koenig@amd.com>
Acked-by: Guchun Chen <guchun.chen@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
On GFXIP9.4.3 APU, allow the memory reporting as per the ttm pages
limit in NPS1 mode.
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
To allow bigger allocations specially on systems such as GFXIP 9.4.3
that use GTT memory for VRAM allocations, relax the limits to
maximize ROCm allocations.
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Needed to avoid a hardware issue.
v2: force high for all GC11 parts for consistency (Alex)
v3: rebase
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
We normally place GART based on the location of VRAM and the
available address space around that, but provide an option
to force a particular location for hardware that needs it.
v2: Switch to passing the placement via parameter
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
On some systems with Navi3x dGPU will attempt to use BACO for runtime
PM but fails to resume properly. This is because on these systems
the root port goes into D3cold which is incompatible with BACO.
This happens because in this case dGPU is connected to a bridge between
root port which causes BOCO detection logic to fail. Fix the intent of
the logic by looking at root port, not the immediate upstream bridge for
_PR3.
Cc: stable@vger.kernel.org
Suggested-by: Jun Ma <Jun.Ma2@amd.com>
Tested-by: David Perry <David.Perry@amd.com>
Fixes: b10c1c5b3a ("drm/amdgpu: add check for ACPI power resources")
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[WHAT]
Add a new field to keep track whether a crtc is previously
writeback-enabled.
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[WHAT]
Handle writeback requests and fill in the required information for DWB
programming and setup.
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
I think this was an abstraction back from when
kfd supported both radeon and amdgpu. Since we just
support amdgpu now, there is no more need for this and
we can use the amdgpu structures directly.
This also avoids having the kfd_cu_info structures on
the stack when inlining which can blow up the stack.
Cc: Arnd Bergmann <arnd@kernel.org>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Increase the retry loops and replace the constant number with macro.
Signed-off-by: Tao Zhou <tao.zhou1@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
No need to perform the full reset operation in case of gpu reset
failure.
Signed-off-by: Tao Zhou <tao.zhou1@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
As part of IP discovery early_init is run for all HW IP blocks.
During this phase all firmware is supposed to be identified that may
be missing so that the driver can avoid releasing resources used by
the EFI framebuffer or simpledrm until the last possible moment.
Move microcode loading from sw_init to early_init.
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
As part of IP discovery early_init is run for all HW IP blocks.
During this phase all firmware is supposed to be identified that may
be missing so that the driver can avoid releasing resources used by
the EFI framebuffer or simpledrm until the last possible moment.
Move microcode loading from sw_init to early_init.
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
As part of IP discovery early_init is run for all HW IP blocks.
During this phase all firmware is supposed to be identified that may
be missing so that the driver can avoid releasing resources used by
the EFI framebuffer or simpledrm until the last possible moment.
Move microcode loading from sw_init to early_init.
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
As part of IP discovery early_init is run for all HW IP blocks.
During this phase all firmware is supposed to be identified that may
be missing so that the driver can avoid releasing resources used by
the EFI framebuffer or simpledrm until the last possible moment.
Move microcode loading from sw_init to early_init.
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
As part of IP discovery early_init is run for all HW IP blocks.
During this phase all firmware is supposed to be identified that may
be missing so that the driver can avoid releasing resources used by
the EFI framebuffer or simpledrm until the last possible moment.
Move microcode loading from sw_init to early_init.
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
As part of IP discovery early_init is run for all HW IP blocks.
During this phase all firmware is supposed to be identified that may
be missing so that the driver can avoid releasing resources used by
the EFI framebuffer or simpledrm until the last possible moment.
Move microcode loading from sw_init to early_init.
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The error path for SDMA firmware loading is unnecessarily noisy.
When a firmware is missing 3 errors show up:
```
amdgpu 0000:07:00.0: Direct firmware load for amdgpu/green_sardine_sdma.bin failed with error -2
[drm:sdma_v4_0_early_init [amdgpu]] *ERROR* Failed to load sdma firmware!
[drm:amdgpu_device_init [amdgpu]] *ERROR* early_init of IP block <sdma_v4_0> failed -19
```
The error code for the device init is bubbled up already, remove the
second one.
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Replace with set_plpd_mode uniformly for places to use.
Signed-off-by: Le Ma <le.ma@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
AGP aperture is deprecated and no longer functional.
v2: fix typo (Alex)
v3: just skip the agp setup call
v4: revert back to the original model
v5: back to v3
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
err_event_athub will corrupt VCPU buffer and not good to
be restored in amdgpu_vcn_resume() and in this case
the VCPU buffer needs to be cleared for VCN firmware to
work properly.
Acked-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: David (Ming Qiang) Wu <David.Wu3@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Fix a memory leak in amdgpu_fru_get_product_info().
Cc: Alex Deucher <Alexander.Deucher@amd.com>
Reported-by: Yang Wang <kevinyang.wang@amd.com>
Fixes: 0dbf2c5626 ("drm/amdgpu: Interpret IPMI data for product information (v2)")
Signed-off-by: Luben Tuikov <luben.tuikov@amd.com>
Reviewed-by: Alex Deucher <Alexander.Deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
To disable AGP, the start needs to be set to a higher
value than the end. Set a default disable value for
the AGP aperture and allow the IP specific GMC code
to enable it selectively be calling amdgpu_gmc_agp_location().
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The BOT register needs to be larger than the TOP register
for this to be properly disabled. The lower 22 bits
of the BOT address are always 0 and the lower 22 bits of
the TOP register are always 1 so you need to make
the upper bits of BOT larger than the upper bits of BOT.
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Yang Wang <kevinyang.wang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This identifies the physical ordering of devices in the hive
v2: fix compilation issue
Signed-off-by: Mangesh Gadre <Mangesh.Gadre@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Otherwise GPU may access the stale mapping and generate IOMMU
IO_PAGE_FAULT.
Move this to inside p->mutex to prevent multiple threads mapping and
unmapping concurrently race condition.
After kfd_mem_dmaunmap_attachment is removed from unmap_bo_from_gpuvm,
kfd_mem_dmaunmap_attachment is called if failed to map to GPUs, and
before free the mem attachment in case failed to unmap from GPUs.
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
For the PASID flushing we already handled that at a higher layer, apply
those workarounds to the standard flush as well.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Instead of each implementation doing this more or less correctly
move taking the reset lock at a higher level.
v2: fix typo
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
That function never fails, drop the error return.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The same PASID can be used by more than one VMID, reset each of them.
Use the common KIQ handling.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The same PASID can be used by more than one VMID, reset each of them.
Use the common KIQ handling.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Testing for reset is pointless since the reset can start right after the
test.
The same PASID can be used by more than one VMID, invalidate each of them.
Move the KIQ and all the workaround handling into common GMC code.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Testing for reset is pointless since the reset can start right after the
test. Grab the reset semaphore instead.
The same PASID can be used by more than once VMID, build a mask of VMIDs
to invalidate instead of just restting the first one.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Shashank Sharma <shashank.sharma@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Testing for reset is pointless since the reset can start right after the
test. Grab the reset semaphore instead.
The same PASID can be used by more than once VMID, build a mask of VMIDs
to invalidate instead of just restting the first one.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Shashank Sharma <shashank.sharma@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Remove leftovers from copying this from the gmc v10 code.
v2: squash in fix from Yifan
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Move the SDMA workaround necessary for Navi 1x into a higher layer.
v2: use dev_err
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The amdgpu_ras_eeprom_control.bad_channel_bitmap is u32 type, but the
channel index could be larger than 32. For the ASICs whose channel
number is more than 32, the amdgpu_dpm_send_hbm_bad_channel_flag
interface is not supported, so we simply bypass channel bitmap update under
this condition.
v2: replace sizeof with BITS_PER_TYPE, we should check bit number
instead of byte number.
Signed-off-by: Tao Zhou <tao.zhou1@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Prepare for bad page retirement.
Signed-off-by: Tao Zhou <tao.zhou1@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If retry cam enabled, we don't use sw retry fault filter and add fault
into sw filter ring, so we shouldn't remove fault from sw filter.
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The KIQ code path was ignoring the second flush. Also avoid long lines and
re-calculating the register offsets over and over again.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
On GFX v9.4.3 dGPU, applications have random timeout failure when XNACK
on, dmesg log has "amdgpu: IH soft ring buffer overflow 0x900, 0x900",
because dGPU mode has 272 cam entries. After increasing IH soft ring
to 512 entries, no more IH soft ring overflow message and application
passed.
Fixes: bf80d34b6c ("drm/amdgpu: Increase soft IH ring size")
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
On a full device reset, PSP FW gets unloaded. Hence restore the
partition mode by placing a new request.
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Asad Kamal <asad.kamal@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Tested-by: Asad Kamal <asad.kamal@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Include subrevision and variant fileds also to IP version.
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Print channel index for UMC v12.
Signed-off-by: Tao Zhou <tao.zhou1@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
update the query to return the number of functional
instances where there is more than an instance of the requested
type and for others continue to return one.
v2: count must reflect the actual number of engines (Alex)
v3: fix wrong number of engines for vcn (Alex)
Signed-off-by: Sathishkumar S <sathishkumar.sundararaju@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
It should first check block ras obj whether be set, it should
return 0 directly if block ras obj or hw_ops is not set.
If block doesn't support RAS just return 0 is fine.
Changed from V1:
return 0 directly if block ras obj or hw ops is not set
Signed-off-by: Stanley.Yang <Stanley.Yang@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
